Review of AnyLogic, Simulation and Digital Twin Software Vendor

By Léon Levinas-Ménard

Last updated: April, 2025

AnyLogic is a comprehensive simulation and digital twin platform that enables organizations to model complex operational processes across sectors such as supply chain, manufacturing, and healthcare. The platform combines three major simulation methodologies—agent‐based, discrete event, and system dynamics—to create dynamic, detailed representations of real‑world systems. By integrating live and historical data into these digital twins, AnyLogic supports robust what‑if analyses and scenario testing without risking actual operations. In addition, the solution leverages external machine learning integrations (for example, via H2O.ai) to predict key performance parameters and forecast capacity needs, while its cloud‑based deployment options facilitate scalable, collaborative execution and interactive dashboard reporting. Built on a Java‑based architecture with extensive API support and customization via user‑provided code, AnyLogic empowers decision makers to explore and optimize process performance in a data‑driven environment.

1. What Does AnyLogic’s Software Deliver?

1.1 Simulation Modeling Capabilities

AnyLogic provides a simulation modeling environment that supports a tri‑modal approach:

Agent‑Based Modeling: Constructs models wherein individual entities (agents) exhibit independent behavior and interact dynamically.
Discrete Event Simulation: Employs process flowcharts to model operations where distinct events occur at specific points in time.
System Dynamics: Captures aggregate flows using stocks and flows to represent continuous processes.

This multimethod capability enables users to select the most appropriate technique—or to integrate methods within a single model—for capturing the nuances of complex real‑world processes ¹².

1.2 Digital Twin Development

AnyLogic positions its solution as a tool for digital twin creation. A digital twin developed on the platform typically includes:

A detailed simulation model that mirrors a physical system’s operating processes (for example, a hospital’s patient flow as demonstrated in a case study ³).
The integration of live or historical operational data via external data feeds, enabling real‑time “what‑if” analyses and scenario testing.
Customizable interactive dashboards and export options (e.g. CSV or Excel) that support managerial decision making.

1.3 Machine Learning and AI Integration

To augment simulation outputs, AnyLogic has incorporated several AI/ML components:

H2O.ai Integration: The platform allows users to incorporate machine learning models—exported as MOJO scoring pipelines—to predict numeric outcomes such as capacity needs ⁴.
Additional Libraries: Tools such as Pypeline, ONNX Helper, and Alpyne are available to bridge simulation models with external ML workflows. In these cases, AnyLogic leverages “black‑box” ML models to supplement rather than replace its core simulation logic.

1.4 Cloud Deployment

AnyLogic offers both public and private cloud solutions for simulation deployment:

Simulation experiments can be executed in parallel via AnyLogic Cloud, with support for RESTful API integration across languages like JavaScript, Python, and Java ⁵.
Models can be shared, run remotely, and export detailed experimental data, all without requiring client‑side installations.
Although powerful, users must configure their models for cloud operation rather than relying on fully automated deployment.

2. How Does AnyLogic’s Solution Work?

2.1 Technical Foundations

AnyLogic’s core environment is built upon Java SE as an Eclipse‑based application. This foundation supports cross‑platform compatibility and an object‑oriented framework that users can extend through custom Java code. Modeling constructs include:

Flowcharts and process blocks for discrete‑event simulations.
Statecharts and agent behaviors for developing agent‑based models.
Stocks, flows, and differential equations for system dynamics models.

Such flexibility empowers users to tailor simulations for complex logistics, manufacturing workflows, or healthcare patient routing ¹⁶.

2.2 Digital Twin Construction

Constructing a digital twin on AnyLogic involves:

Building a simulation model that accurately reflects a physical system’s workflow.
Dynamically linking the model with operational databases or real‑time data feeds.
Capturing key performance metrics (like waiting times and bed utilization in a hospital setting) that can be continuously compared with actual data for validation and improvement ³.

2.3 AI/ML Integration Implementation

AnyLogic integrates external ML capabilities in a modular fashion:

Pre‑trained machine learning models (e.g. from H2O.ai) are exported as standalone files and “called” from within the simulation. This enables predictions such as patient length‑of‑stay or production rates.
The simulation remains the core decision‑support tool, with ML predictions supplementing the primary discrete simulation logic ⁴.

2.4 Cloud and Deployment Mechanisms

The AnyLogic Cloud is engineered to run simulations in the background while delivering interactive animations and dashboards via modern web browsers:

A load‑balancing system reuses results for identical input configurations to conserve compute time.
Custom APIs allow integration with larger enterprise workflows and support custom frontend development ⁵.

3. Evaluating the State-of-the-Art

3.1 Strengths

Comprehensive Multimethod Simulation: AnyLogic stands out by integrating all three simulation methodologies into one package, a feature well supported by educational resources like “The Big Book of Simulation Modeling” ².
Open APIs and Extensibility: With support for Java, Python, and JavaScript, users can deeply integrate AnyLogic models with external systems and adapt them for diverse applications.
Cloud‑Enabled Deployment: The scalable, collaborative environment of AnyLogic Cloud enhances both research and real‑time operational analysis.

3.2 Points for Skepticism

AI Claims: While marketed as “AI‑enabled,” the core artificial intelligence functionality relies on third‑party integrations rather than an intrinsic deep‑learning engine.
Digital Twin Complexity: The creation of accurate digital twins demands significant domain expertise and careful data integration, meaning success is highly dependent on the quality of the underlying models and data.
Incremental Improvements: Although cloud‑enabled features and model reusability offer operational benefits, these improvements may be evolutionary rather than revolutionary relative to other simulation or optimization platforms.

AnyLogic vs Lokad

AnyLogic and Lokad represent two distinct approaches in the realm of supply chain and operational decision support. AnyLogic focuses on sophisticated simulation and digital twin construction; it enables users to replicate real‑world processes through agent‑based, discrete event, and system dynamics modeling, thereby offering a flexible environment for scenario analysis and what‑if testing ¹³. In contrast, Lokad centers on quantitative supply chain optimization through predictive decision making. It features a purpose‑built platform with a proprietary DSL (Envision) and integrated machine learning engines designed to deliver concrete recommendations—such as precise inventory or pricing actions—on a daily basis ⁷⁸. Architecturally, AnyLogic is rooted in a Java‑based, open‑integration environment ideal for customizable simulations, whereas Lokad leverages F# and C# in a tightly integrated cloud‑hosted solution that minimizes third‑party dependencies ⁸⁹. Ultimately, while AnyLogic is best suited for organizations seeking to explore dynamic operational scenarios and build digital twins, Lokad offers a more prescriptive, automation‑oriented platform aimed at directly optimizing supply chain decisions.

Conclusion

AnyLogic delivers a robust and versatile simulation platform that empowers organizations to create detailed digital twins and model complex systems for informed decision making. Its strength lies in offering a comprehensive, multimethod simulation environment combined with cloud‑based collaboration and external ML integrations. Nevertheless, the platform’s reliance on third‑party AI components and the resource‑intensive process of developing accurate digital twins require significant expertise and careful implementation. When compared with platforms like Lokad—whose tightly integrated, optimization‑driven approach provides prescriptive, automated decision support—AnyLogic remains invaluable for simulation‑driven analysis and scenario planning. Organizations must carefully assess their strategic needs and internal capabilities to determine which platform best aligns with their operational objectives.